DebriSolver Competition

Advancing Space Traffic Management through Data-Driven Solutions

The Challenge

As the number of satellites in orbit continues to grow, effective space traffic management has become a global priority. In August and September 2024, two debris-generating events produced over 700 new objects, threatening hundreds of active spacecraft and revealing significant regulatory and technical gaps in collision prevention. Because these satellites support critical services such as telecommunications, Earth observation, navigation, and emergency response, their loss could severely impact global connectivity, security, and economic activity. Addressing these risks requires an integrated approach that combines technological innovation with coordinated policy, operations, and economic planning.

The Team

Elena Ancona

Space Traffic Management Solutions Engineer

STM Solutions Engineer with over 10 years of experience in space operations. PhD in Aerospace Engineering, with professional background including operational roles at ESA and EUMETSAT. Former OneWeb Flight Dynamics Engineer, now collaborating with OKAPI:orbits on space sustainability.

LinkedIn →

Giovanni Facchinetti

Space Debris Mitigation Engineer

Space engineer focused on debris mitigation and orbital sustainability. Double Master's in Space Engineering from Politecnico di Milano and Beihang University. Worked for ESA's Advanced Concepts Team, specializing in long-term debris environment modeling.

LinkedIn →

Vincenzo Quatraro

AI & Data Analytics Consultant

AI and Machine Learning consultant and AI Team Lead at Martur Fompak International, also working freelance as an AI Functional and ML/AI Solutions Architect. Mechanical Engineering graduate, with a Master’s in AI and Machine Learning from the University of Leeds and PhD research in Aerospace Engineering.

LinkedIn →

Diego Guerra

Spacecraft Telecommunications Engineer

Principal Telecommunications Lead at Blue Origin on NASA Artemis Lunar Lander. Previously worked at SpaceX on first-generation Starlink LEO constellation. PhD in Electrical Engineering from Arizona State University.

LinkedIn →

Our Solutions

This study analyzes the 2024 Long-March 6A fragmentation events (700+ fragments) to quantify risks to critical space infrastructure. Through a systematic analysis of over 500.000 CDMs, we identify operational criticalities in current STM frameworks and propose an integrated solution that combines technical innovation, regulatory harmonization (ESA-NASA standards), economic incentives (orbital tolls), and social protection (resilience of critical services). We introduce the STM-ConOps Score, a composite metric for mission classification enabling proportionate regulation and quantitative guidelines compliance assessment.

The following sections summarize our solution at a high level.
Click on each section to explore the underlying methodology in more detail.

Our Approach

Integrated technical, regulatory, and economic solutions for orbital resilience

Data Analysis

Systematic analysis of 570,000+ CDM messages and fragmentation events

Key Findings

Critical insights on debris dynamics and conjunction risk patterns

Results

Visualizations and quantitative assessments of orbital safety challenges

Our Approach

Treating STM as a systemic challenge, we link technical risk detection with policy and economic sustainability for long-term orbital resilience. Our solution integrates technical innovation with regulatory and economic foresight, analyzing 2024 fragmentation events to extract lessons learned and propose actionable recommendations.

Key Innovation: STM-ConOps Score

Composite indicator capturing operational, environmental, and governance externalities across 6 critical dimensions, enabling proportionate regulation and risk-based economic tools.

Orbital Regime Sensitivity (ORS)

Captures both the current and projected congestion of the target orbital shell, conjunction rates, debris lifetime, and proximity to sensitive regions. This dimension highlights the importance of anticipating future environmental stress due to planned launches and constellation growth.

Spacecraft Physical & Operational Characteristics (SPOC)

Assesses trackability, maneuverability, propulsion capability, attitude control reliability, and passivation features. These properties determine the spacecraft's ability to participate safely in the orbital traffic system.

Mission Behaviour Complexity (MBC)

Reflects operations that increase unpredictability, such as formation flying, multi-phase deployments, or proximity operations. Complex missions demand increased coordination and represent a higher potential for operational conflict.

Debris & Failure Externality Potential (DFEP)

Quantifies the consequences of a catastrophic failure, based on mass, fragmentation likelihood, stored energy (e.g., pressurised tanks), number of spacecraft involved, and compliance with post-mission disposal guidelines.

Governance & Coordination Readiness (GCR)

Evaluates adherence to best practices, quality of ephemeris sharing, manoeuvre notifications, anomaly reporting, and transparency. Operational culture is a key determinant of STM safety.

Failure-Mode Severity & Recoverability
(FMSR)

Measures the spacecraft's ability to maintain controllability, enter safe modes, and remain trackable during anomalies. Recoverability is essential for reducing uncontrolled conjunction risk.

Multi-Domain Recommendations

An integrated STM solution combining technical, regulatory, economic, and social dimensions to enhance orbital safety and sustainability.

🔬

Technical Innovation

Enhanced SSA accuracy, systematic conjunction classification, and operator ephemeris prioritization

⚖️

Regulatory Framework

Harmonize international standards, encourage data sharing, and establish unified STM framework

💰

Economic Incentives

Risk-based orbital tolls, insurance mechanisms, and cost-sharing for active debris removal

🛡️

Social Protection

Critical services protection, UNCOPUOS alignment, and fostering international cooperation

Data Analysis

STEP A

Fragment Cloud Evolution

Reconstruct temporal evolution of debris populations by ingesting CDMs and tracking unique NORAD IDs. Binning logic separates "active" vs "inactive" fragments, with intelligent reactivation mechanism.

STEP B

Conjunction Events Analysis

Systematic grouping of CDMs into physical conjunctions (same object pair and TCA). Classification: Critical (CR), Concerning (CN), Nominal (NO) based on collision probability, miss distance, and time-to-TCA.

STEP C

Trend Evaluation & Criticality

Assess CDM source impact (HAC vs EPHEM). Develop maneuver detection logic through multiple CDMs analysis, source alternation patterns, and large Pc/MD shifts. Generate insights on operators' behavior patterns.

STEP D

TLE-Based Maneuver Detection

TLEs (±7 days around TCA) screened for abrupt semi-major axis changes (ΔSMA > 400m) in the most critical conjunctions, validating operational response to high-risk events.

Data Sources

📊

570,000+ CDM Messages

Conjunction Data Messages of the fragmentation cloud

🛰️

Space-Track TLEs

Two-Line Element sets for orbital tracking

📡

Seradata Dataset

Contextualization and satellite information

🌐

Slingshot Laboratory

3D visualization and analysis tools

🐍

Custom Python Routines

Automated data processing and modeling

Key Findings

Systemic Risk Drivers

Event-level CDM classification and debris-cloud dynamics reveal systemic risk drivers in congested orbital regimes.

Persistent Fragmentation Impacts

Fragmentation impacts persist and evolve long after the initial event. Debris populations show continued activity through the CDMs dataset.

Broader Conjunction Context

Conjunctions cannot be treated as isolated events. Effective collision avoidance must account for the broader orbital neighborhood and constellation-level interactions, not just single object pairs.

Results

The figures and tables presented below are interactive. Additional information and details can be accessed by clicking on the individual elements.

TLEs for Manouver Detection

Fragments Cloud Evolution

Fragments Population Dynamics

CDMs Classification

Tables

Conjunction events with the highest critical score, and their respective count of critical CDMs (for CR3 CDMs, the source is also given, if Ephems or HAC).

Conjunctions with the highest number of critical CDMs.

Pairs of Primary/Secondary NORADs with the highest number of events and the total number of CDMs

The average delta time between two consecutive events’ TCAs helps identify potential repeated conjunctions

Pairs of Primary/Secondary NORADs with highest n° of CDM

Pairs of Primary/Secondary NORADs with the highest cumulative critical score

Future work: AI Application

Future work will explore the integration of AI methods to extend the current event-based analysis. The results show that orbital risk is driven by dynamic interactions, making it suitable for models capable of capturing non-linear and temporal patterns.

AI will be applied as a support tool for operators, enhancing risk monitoring across CDM updates, and as a pattern-discovery mechanism to identify relationships not evident through rule-based analysis. The use of Explainable AI (e.g. XGBoost with SHAP) will ensure that insights remain transparent, interpretable, and operationally relevant, while allowing the framework to scale as new variables and data sources are introduced.

AI Models: ideas and strategies

We define a multi-model strategy: XGBoost for operational deployment and SHAP for transparent explainability.

XGBoost (Extreme Gradient Boosting)

Handles non-linear feature interactions natively
Robust to missing data (common in CDM streams)
Scales efficiently to 120.000+ events
Built-in feature importance for interpretability
Industry-proven for operational deployment

🔍 SHAP Explainability

Summary Plot: Global feature importance ranking across all events
Dependence Plot: Feature interactions (e.g., PC slope vs. MD convergence)
Force Plot: Individual event explanations with feature attribution

SHAP Beeswarm Plot – Feature Impact Overview

The plot illustrates the contribution of the five features selected in this initial approach to the model’s predictions. The SHAP value distribution highlights how each feature influences the output across the dataset, revealing variability, non-linear effects, and potential interactions while supporting model interpretability.